Prompt more, merge more,
ship more.

No. Bloom's interface is built with AppKit and SwiftUI. Its browser pane uses a web view to display websites.

Running more agents can use more of your provider's allowance. Usage and costs depend on the tasks, models and plan you use. Bloom adds no charge; your existing subscription or API billing still applies.

Claude Code and Codex, using the CLI and login you already have on your Mac. Cursor and OpenCode are not supported yet.

Yes. Connect your client using the command in Bloom's Settings. You can list projects, create workspaces, start tasks and check their progress over MCP. Agents running inside Bloom can also control the panes in their workspace.

Yes. Bloom is free to download and use, with no paid tier planned. Spatie funds it through our other work. You still need your own coding agent subscription or API access.

Yes. You can read the code, report bugs and contribute at github.com/spatie/bloom.

Each worktree starts as a fresh checkout. Bloom copies your .env files and runs the project's setup script if one is configured. Use that script to install dependencies and set up services. Build caches start cold in each new worktree.

Each workspace starts in a fresh directory. For a Laravel project, you typically need to install dependencies, configure a local URL and prepare a database before you can run the app.

A setup script is a text file containing the terminal commands you would normally run to get your project ready. Bloom runs them for you in each new workspace, before the agent starts. You only need to configure this once per project.

1. Create the script in your project. Open your original project folder in your editor, the one containing artisan and composer.json. Create a folder named .bloom there if it doesn't exist, then create a file inside it named setup.sh. The full path within your project is .bloom/setup.sh.

Put your setup commands in that file. The example below is based on There There's setup and assumes you already use Laravel Valet, PostgreSQL and Bun on your Mac. Replace my-app and my_app with your project's name, and adjust the commands if you use different tools. If you're unsure what belongs here, ask your coding agent to adapt this example to your project.

#!/usr/bin/env bash

# Stop on errors or unset variables.
set -eu

# Bloom provides the original checkout path and this workspace's ID.
# Start with the original project's environment settings.
cp "$BLOOM_ROOT_PATH/.env" .env

# Give this workspace its own HTTPS .test domain with Laravel Valet.
SITE="my-app-$(printf '%s' "$BLOOM_WORKSPACE_ID" | tr -cd '[:alnum:]' | cut -c1-10)"
valet link "$SITE"
valet secure "$SITE"
sed -i '' "s|^APP_URL=.*|APP_URL=https://$SITE.test|" .env

# Use a separate PostgreSQL database so the main checkout is untouched.
# Only create it if it doesn't exist, so setup can run again.
DATABASE="my_app_$(printf '%s' "$BLOOM_WORKSPACE_ID" | tr -cd '[:alnum:]' | cut -c1-12)"
if ! psql -d postgres -tAc "SELECT 1 FROM pg_database WHERE datname = '$DATABASE'" | grep -q 1; then
    createdb "$DATABASE"
fi
sed -i '' "s|^DB_DATABASE=.*|DB_DATABASE=$DATABASE|" .env

# Install PHP and JavaScript dependencies, then build the frontend assets.
composer install
bun install
bun run build

# Create the tables and seed the workspace database with initial data.
php artisan migrate --seed --force

2. Tell Bloom where to find it. In the same .bloom folder, create settings.toml and add the following. If the file already has a [scripts] section, add the setup_file line to that section instead of creating a second one.

[scripts]
setup_file = ".bloom/setup.sh"

3. Make the script runnable and save both files in Git. Open a terminal in your original project folder and run these commands. Committing the files makes them available in new workspaces. If you're doing this on a feature branch, merge it into the branch you create workspaces from.

chmod +x .bloom/setup.sh
git add .bloom/setup.sh .bloom/settings.toml
git commit -m "Add Bloom setup script"

4. Create a new workspace in Bloom. Bloom runs the script from the new workspace's folder and supplies the BLOOM_* variables used above. If setup fails, check its output, fix the script and choose Run Setup Again from the Workspace menu.

Once setup works, you can also create .bloom/archive.sh to remove the workspace's Valet site and database when you archive it. Register it with archive_file = ".bloom/archive.sh" in the same [scripts] section. Keep cleanup limited to the URL and database created for that workspace.

Quitting Bloom stops running agent turns. To keep terminals running, install tmux and turn on "Keep terminals running after quitting" in Settings. Those terminals will still be available when you reopen Bloom.

No. Each workspace is a git worktree on your disk. You can open it in your editor, run your own tools and push its branch as usual.

No. We have no plans for Windows or Linux.